- Fitness History: based on the scores computed in
past evolutionary processes. The Fitness Graph displays the progress of each ecosystem.
- Score Diversity: used to determine the fitness of
each entity in the ecosystem population. The score diversity plot shows the distribution
of scores in each population.
- Truth and Prediction: The output values are
calculated from each row of data and then compared to the training table data.
"True" values are then plotted against "Calculated" values to display
the progress of each evolutionary pass.
- Output Code: Derived solutions are
straightforward, understandable, and in a very useable format of pseudo code, C, FORTRAN,
or Basic.
What is an Evolutionary Algorithm? e
uses a process modeled after biological evolution to evolve computer programs. Randomly
generated populations of computer code which represent candidate solutions to your problem
are evaluated for fitness and then selected to parent the next improved generation of
programs. For example, use e to
find a signal processing algorithm, to encapsulate a calibration table, to discover a
securities buy/sell rule, or to evolve a decision algorithm.
Visit the web site of System Dynamics International (SDI). This
high technology engineering services firm has designed this advanced Evolutionary
Algorithm Tool to solve very difficult problems with software for the engineering and
scientific industries.
Extensive evaluations made by QMC found e
Model to be the highest quality program available. A comparison for the results of three
modeling approaches for a polymerization process is available.
KEY BENEFITS
- e minimizes
preparation effort and optimizes your time
- Provides the actual equations
- No initial assumptions are needed
Use e whenever
you need to discover the relationship (algorithm, model, control law, etc.) between
variables and a desired output. e Model
has been successfully used for modeling systems with over nine variables and provides the
actual equation to model the process. Results for a polymer plant process show a relative
error and standard error deviation reduction of 50% when compared to neural networks using
the same data for modeling, testing, and on-line prediction.
KEY FEATURES
- Model is evolved at solution time
- Variety of solutions are explored
- Computer software searches for solution
- Solution is as good as empirical data collected
- Solutions are easily adapted to changing requirements
- Libraries of past learning are used for new problems.
- Outputs to C, FORTRAN, or Basic
USE FOR
- Pattern Recognition
- Numerical Analysis
- Sensor Fusion
- Component Response
- Economics
- Engineering
- Signal Processing
- Mathematics
- System Response
- Control and others
SYSTEM REQUIREMENTS
- Computer:
- Processor/Speed 486/66 MHz, IBM Compatible PC
- Memory:
- 8 MB RAM
- Disk Space:
- 8 MB
- Operating System:
- Windows 31, 95, NT
|